LinkedIn is 41% AI slop. A real voice is now the moat.

Here's a number that should change how you think about your marketing: 41% of long-form LinkedIn posts are AI-generated. That's the finding from a new Pangram study measuring machine-written text across five platforms, and LinkedIn is the runaway leader. The takeaway for anyone publishing to sell — a newsletter, a blog, a company page — isn't "AI writing is bad." It's that generic AI content is now the baseline noise, and standing out requires the one thing a model can't fake: a real voice from someone who's actually done the work.

What actually happened

Pangram, which builds an AI-text detector, analyzed roughly a million posts its users organically scrolled across LinkedIn, Medium, X, Reddit, and Substack over a two-month window. As The Decoder and Fast Company reported, LinkedIn led every platform: 41% of its long-form posts (over 250 words) were flagged as AI-generated, with only about 55% written by a human unassisted.

Across all five platforms combined, roughly one in four long-form posts was machine-written. LinkedIn made up about a third of the posts scanned but accounted for nearly two-thirds of all the AI content detected. X wasn't far behind — a quarter fully AI-authored, another ~23% AI-assisted. Substack came out cleanest at around 10%. The picture is consistent: the feeds where people go to build a professional reputation are the ones filling fastest with text nobody actually wrote.

Why it matters for your business

If a quarter of the content in your channel is machine-generated, the median post is now indistinguishable mush — same three-bullet structure, same "Here's why that matters," same confident nothing. Buyers have learned to scroll past it. Which means the bar to earn attention didn't rise because AI got good; it rose because AI made average free and infinite. Competing on volume is a race to the bottom of a feed that's already drowning.

The edge is specificity only an operator has: the real number from the job you ran, the mistake you actually made, the opinion a marketing committee would've sanded off. That's not anti-AI — use the tools to draft, research, and edit faster. It's about what goes in: lived detail a model can't invent because it wasn't there. And the structural move underneath it — own the channel where that voice lives. Rented reach on a slop-saturated feed is a weak foundation; an email list and a site you control are an audience that's actually yours. We write like people who've run a business because we have, and we build the owned channels — site, list, content engine — that make a real voice compound instead of evaporating into someone else's algorithm.

Sources: The Decoder, Fast Company.